[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference rusure::math

Title:Mathematics at DEC
Moderator:RUSURE::EDP
Created:Mon Feb 03 1986
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:2083
Total number of notes:14613

1904.0. "Concours General, France, 1992, problem 4" by EVTSG8::ESANU (Au temps pour moi) Thu Oct 13 1994 07:56

Let u(n) be a sequence of real numbers, defined by

	0 < u(0) < 1
	0 < u(1) < 1
		 (sqrt(u(n)) + sqrt(u(n-1))
	u(n+1) = --------------------------
			     2

a) Prove that the sequence u(n) is convergent.
b) Prove that, beginning with a certain term n0, the sequence (u(n))n>=n0
is strictly increasing.


(The "Concours General" is the French Mathematics Olympiad).


Mihai.
T.RTitleUserPersonal
Name
DateLines
1904.1A solutionBALZAC::QUENIVETMargins are often too small.Mon Dec 12 1994 08:4654
Here is my solution, proving b) first, then a).

1. For each n, u_n is well defined and 0 < u_n < 1.

	This is true for n = 0 and n = 1 by definition. And by induction 
we have for n >= 2:

	u_n = (sqrt(u_(n-1)) + sqrt(u_(n-2))/2 > 0
	and
	u_n =  (sqrt(u_(n-1)) + sqrt(u_(n-2))/2 < (sqrt(1) + sqrt(1))/2 = 1 

2. The sequence u_n is (strictly) increasing.

	Let's compute for an n >= 2:

	u_n - u_(n-1) = - u_(n-1) + 1/2 sqrt(u_(n-1)) + 1/2 sqrt(u_(n-2))

	In order to know the sign of this expression, we have to look at
	the following function (for a fixed n and a variable x):

	x |--> f(x) = -x� + 1/2 x + 1/2 sqrt(u_(n-2))

	The roots of the equation f(x) = 0 are:

	x_1 =  ( 1 - sqrt( 1 + 8 * sqrt(u_(n-2) ) ) / 4

	x_2 =  ( 1 + sqrt( 1 + 8 * sqrt(u_(n-2) ) ) / 4

	For each x in the interval [x_1, x_2], we have f(x) > 0 (the opposite
sign of the leading coefficient) and f(x) < 0 for each x outside this interval.

Since  f(0) = 1/2 sqrt(u_(n-2)) > 0 and 
       f(1) = 1/2(1+sqrt(u_(n-2)) > 0, we conclude that f(x) > 0 for each
x in [0, 1]. (In other words, the interval [0, 1] is included in 
the interval [x_1, x_2]).

Since u_(n-1) is in [0, 1], also sqrt(u_(n-1)) is in [0, 1] and

u_n - u_(n-1) = f( sqrt ( u_(n-1) ) ) > 0. 

3. The real sequence u_n has an upper bound ( =1 ) and is increasing.
Then it has a limit l. For continuity reasons (of the square root function), 
this limit must verify:

l = 1/2(sqrt(l) + sqrt(l)), or l = sqrt(l), so l = 0 or l = 1. 

Since u_n >= u_2 > 0 for each n >= 2, l > 0, so l = 1.


I don't see a simple way to prove a) first then b).

Herv�

    
1904.2re .1: Retry...EVTSG8::ESANUAu temps pour moiMon Dec 12 1994 09:1410
>       f(1) = 1/2(1+sqrt(u_(n-2)) > 0,

Herve, you made a slight mistake:
f(1) =  (1/2)(-1 + sqrt(u_(n-2)) < 0

So...

I found no way of proving b) before a).

Mihai.
1904.3RTL::GILBERTMon Dec 12 1994 15:3312
>2. The sequence u_n is (strictly) increasing.

I'm suprised there's no simpler way of proving this (and I've looked).

>3. The real sequence u_n has an upper bound ( =1 ) and is increasing.
>Then it has a limit l.

This isn't clear.  For example, the sequence:
	v_1 = 0, v_2 = 1/4, v_n = v_<n-1> + 2^(-n)
has an upper bound of 1 and is increasing.  But that doesn't make its limit 1.

(FWIW, the v sequence's _least_ upper bound is 1/2, as is its limit).
1904.4Let's put it in other wordsBALZAC::QUENIVETMargins are often too small.Tue Dec 13 1994 03:5821
    Let's make it clear: I said (and it's true) that a sequence which has
    an upper bound (in this case the real number 1, one) and is increasing
    has a limit (let's call it l, the letter 'l' for 'Linda', 'Louise' or
    whatever you like). 
    
    Since the bound is 1 (one) all we can say is that l
    (letter) is less or equal to 1. But the limit of u_(n-1) is l, also the
    limit of sqrt(u_(n-1)) is sqrt(l), and same for u_(n-2). So the limit
    of 1/2(sqrt(u_(n-1))+sqrt(u_(n-2)) is 1/2(sqrt(l)+sqrt(l))=sqrt(l).
    
    But 1/2(sqrt(u_(n-1)+sqrt(u_(n-2)) is u_n by definition, whose limit is
    l (for 'Laurence': it's a female name in French), so l = sqrt(l) and
    then l is 0 or 1. It cannot be 0 since all the terms of the sequence
    are greater than the term u_2 which is *strictly* greater than 0.
    
    Herv�
    
    BTW, I feel a little ashamed of the wrong proof in .1: I wasted my
    week-end :-)
    
     
1904.5AUSSIE::GARSONachtentachtig kacheltjesTue Dec 13 1994 20:404
    re .4
    
    I too misread the l as a 1. I would humbly request that l not be used
    in proofs. Anyway .4 clears it up.
1904.6.0 has not been proved,EVTSG8::ESANUAu temps pour moiThu Dec 15 1994 08:311
and it is most enjoyable!